Permutation Symmetry and the Exchange Operator

The previous article left us with a crisis: for identical particles, the state $\psi(1,2)$ , its relabeled twin $\psi(2,1)$ , and every superposition between them predict exactly the same physics — an exchange degeneracy the formalism cannot tolerate. This article builds the machinery that resolves it, and proves the single most surprising fact in the whole subject: nature uses only two symmetry classes, no matter how many particles are involved. We start by making the label-swap an operator, $\hat P_{12}$ , and reading its structure straight off $\hat P_{12}^2=\mathbb 1$ : it is both Hermitian and unitary, with eigenvalues exactly $\pm 1$ . Because a genuine identical-particle Hamiltonian is symmetric in its labels, $[\hat P_{ij},\hat H]=0$ , so exchange character is a conserved quantum number — a boson can never drift fermionic. We build the symmetrizer $\hat S$ and antisymmetrizer $\hat A$ and prove their projector algebra. Then comes the owned theorem. For two particles the story looks complete: symmetric plus antisymmetric spans everything. For three or more it is not, and the reason is sharp and computable — the transpositions stop commuting, $[\hat P_{12},\hat P_{23}]\neq 0$ , so there is no basis in which every pair-swap has a simultaneous definite sign except the totally symmetric and totally antisymmetric sectors. A machine dimension census makes it concrete: three qubits split $8=4+4+0$ , and the antisymmetric dimension is identically zero — three spin- $\tfrac12$ fermions cannot all fit in two one-particle states. That is Pauli exclusion, appearing here before we have named it. What we do not do: build the explicit two-particle $\pm$ states with normalization (next article), assign the sign to each species, or develop the $N$ -fermion Slater determinant. Those are named and deferred; here we own the operator algebra and the two-sector theorem.

⚡ Key Takeaways

The exchange operator $\hat P_{ij}$ swaps two particles' labels. $\hat P_{ij}^2=\mathbb 1$ makes it Hermitian and unitary, with eigenvalues exactly $\pm 1$ : symmetric and antisymmetric eigenspaces.
Identical particles force $[\hat P_{ij},\hat H]=0$ , so exchange parity is conserved — the symmetry class of a state is a constant of motion, and the boson/fermion sectors are dynamically sealed off from each other.
The symmetrizer $\hat S=\tfrac1{N!}\sum_P \hat P$ and antisymmetrizer $\hat A=\tfrac1{N!}\sum_P \text{sgn}(P)\hat P$ are orthogonal projectors: $\hat S^2=\hat S$ , $\hat A^2=\hat A$ , $\hat S\hat A=0$ .
The owned theorem: for $N\ge 3$ the transpositions do not commute ( $[\hat P_{12},\hat P_{23}]\neq 0$ ), so only the totally symmetric and totally antisymmetric sectors give every pair-swap a definite sign. The symmetrization postulate is the empirical statement that nature populates exactly these two.
A dimension census makes it physical: $(\mathbb C^d)^{\otimes 3}$ splits as $\tfrac{d(d+1)(d+2)}{6}$ symmetric, a mixed middle, and $\tfrac{d(d-1)(d-2)}{6}$ antisymmetric. At $d=2$ the antisymmetric dimension is zero — exclusion, before exclusion.
KEY IDEA: exchange symmetry is an operator algebra whose non-abelian structure for $N\ge 3$ selects exactly two physical sectors — the deep reason the world is made of bosons and fermions and nothing in between.

💡 The Core Insight

Permutation symmetry is not a footnote but an operator algebra. The exchange operators square to the identity, giving $\pm 1$ eigenvalues; they commute with any identical-particle Hamiltonian, making exchange parity conserved; and for three or more particles they fail to commute with each other, which is precisely why nature admits only the totally symmetric and totally antisymmetric states — bosons and fermions, and no third kind.

The physical picture

Indistinguishability told us that a physical state must be an eigenstate of every label-swap. This article asks the operator questions that answer how and how many: what does the swap operator look like, what are its eigenvalues, does it survive time evolution, and — the deep one — can we consistently demand a definite swap-sign for every pair at once when there are three or more particles? The two-particle answer is deceptively clean and hides the real subtlety. With one pair there is one swap operator, its square is the identity, and the state space splits perfectly into a symmetric half and an antisymmetric half. Nothing warns you that this completeness is special to $N=2$ . The three-particle answer breaks the illusion. Now there are three pair-swaps, and swapping $1\!\leftrightarrow\!2$ then $2\!\leftrightarrow\!3$ is not the same as doing them in the other order — the operators do not commute. You cannot label a state by a separate sign for each pair unless those signs are all $+1$ (totally symmetric) or all $-1$ (totally antisymmetric). Everything else is "mixed symmetry," a genuinely multidimensional zoo that quantum mechanics simply does not use for fundamental particles. The plot below is that census, computed: the split of three identical objects among $d$ single-particle states, with the antisymmetric bar highlighted where it collapses to zero.

Figure

The dimension census — three particles, d modes, and the sector that vanishes

The census is not decoration; it is the theorem in numbers. The mixed sector is real, has integer dimension, and grows fast — yet no fundamental particle ever lives there. Why the world discards a whole sector is the empirical content of the symmetrization postulate, and the operator reason it is allowed to is the non-commutation we prove below.

The exchange operator and its algebra

Define the operator that does the swapping. It acts on the two-particle tensor product by interchanging the contents of the two slots.

\hat P_{12}\,|\alpha\rangle_1|\beta\rangle_2 = |\beta\rangle_1|\alpha\rangle_2,

where: $|\alpha\rangle_1|\beta\rangle_2$ means particle in slot 1 is in state $\alpha$ , slot 2 in state $\beta$ ; $\hat P_{12}$ trades the two. Applied to a general wavefunction, $\hat P_{12}\psi(\xi_1,\xi_2)=\psi(\xi_2,\xi_1)$ . Its defining structural fact follows from swapping twice returning the original.

\hat P_{12}^2=\mathbb 1 \;\Rightarrow\; \hat P_{12}=\hat P_{12}^{-1}, \qquad \hat P_{12}=\hat P_{12}^\dagger.

where: the first identity says $\hat P_{12}$ is its own inverse (unitary), and — because permutation matrices in a real basis are symmetric — it is also Hermitian. An operator that is simultaneously Hermitian and unitary can only have eigenvalues on the unit circle and real, so:

\hat P_{12}\,|\psi_\pm\rangle = \pm|\psi_\pm\rangle,

where: $|\psi_+\rangle$ (eigenvalue $+1$ ) are symmetric states and $|\psi_-\rangle$ (eigenvalue $-1$ ) antisymmetric. For a $d$ -dimensional single-particle space the symmetric eigenspace has dimension $\tfrac{d(d+1)}{2}$ and the antisymmetric $\tfrac{d(d-1)}{2}$ ; for $d=3$ that is $6$ and $3$ , verified numerically. Because the two particles are genuinely identical, the Hamiltonian is unchanged by relabeling them, which makes exchange a symmetry in the dynamical sense.

[\hat P_{ij},\hat H]=0 \;\Rightarrow\; \frac{d}{dt}\langle\hat P_{ij}\rangle=0,

where: a Hamiltonian symmetric in its labels — kinetic terms $\tfrac{p_1^2+p_2^2}{2m}$ , symmetric couplings $V(|\hat r_1-\hat r_2|)$ , identical external potentials — commutes with every $\hat P_{ij}$ . Under the evolution $e^{-i\hat Ht/\hbar}$ the symmetry class is therefore frozen: a state prepared symmetric stays symmetric forever. This is why "boson" and "fermion" are permanent labels, not phases a particle can slip between.

Derivation AConservation of exchange parity and the projector algebra

This card establishes two facts the rest of the chapter leans on: that exchange parity is a genuine constant of motion for identical particles, and that the symmetrizer and antisymmetrizer are orthogonal projectors. Both are short, and both are checked symbolically for $N=3$ in the verification script. Watch that conservation needs the Hamiltonian to be label-symmetric — the physical input, not a mathematical accident.

Step 1. Conservation. For identical particles the Hamiltonian satisfies $\hat P_{ij}\hat H\hat P_{ij}^{-1}=\hat H$ , i.e. $[\hat P_{ij},\hat H]=0$ . By the Ehrenfest relation for a conserved operator, $\tfrac{d}{dt}\langle\hat P_{ij}\rangle = \tfrac{i}{\hbar}\langle[\hat H,\hat P_{ij}]\rangle = 0$ . The symmetry character is a good quantum number, sealed under $e^{-i\hat Ht/\hbar}$ — the verification script evolves a symmetric two-particle state under a random symmetric $\hat H$ and finds it stays symmetric to $10^{-12}$ , while an artificially label-asymmetric $\hat H$ breaks the sector.

Step 2. Build the projectors. Define $\hat S=\tfrac1{N!}\sum_{P\in S_N}\hat P$ and $\hat A=\tfrac1{N!}\sum_{P\in S_N}\text{sgn}(P)\,\hat P$ , sums over all $N!$ permutations. Each $\hat S|\Psi\rangle$ is invariant under any further permutation (relabeling the sum just reorders it), so $\hat S$ lands in the totally symmetric subspace, and likewise $\hat A$ in the antisymmetric one.

Step 3. Idempotence and orthogonality. Because permutations form a group, $\hat P\hat S=\hat S$ for every $P$ , hence $\hat S^2=\tfrac1{N!}\sum_P \hat P\hat S=\hat S$ . The signed version gives $\hat A^2=\hat A$ . And $\hat S\hat A=0$ : applying any transposition leaves $\hat S$ fixed but flips the sign inside $\hat A$ , so their product must vanish. All three identities are confirmed symbolically for $N=3$ on $(\mathbb C^2)^{\otimes 3}$ .

Step 4. The $N=2$ special case. With two particles there is exactly one non-trivial permutation, and $\hat S+\hat A=\tfrac12(1+\hat P_{12})+\tfrac12(1-\hat P_{12})=\mathbb 1$ : the symmetric and antisymmetric subspaces exhaust the whole space. This is the completeness that looks like the end of the story — and is not, as Step 4's failure for $N=3$ (where $\hat S+\hat A\neq\mathbb 1$ ) reveals below. Why it matters: conservation is what makes the statistics permanent, and the projectors are the tools that build symmetric and antisymmetric states in the next article. The $N=2$ exhaustiveness is the trap that makes the two-sector rule look automatic. Verified: the script confirms $[\hat P,\hat H]=0$ with sector-preserving time evolution (residual $10^{-12}$ ), the projector algebra $\hat S^2=\hat S,\hat A^2=\hat A,\hat S\hat A=0$ for $N=3$ , and $\hat S+\hat A=\mathbb 1$ for $N=2$ but not for $N=3$ .

The owned theorem: why only two sectors

Here is the result this article contributes to the chapter and to no other. For two particles the symmetric/antisymmetric split is exhaustive, and it is tempting to conclude that "symmetric or antisymmetric" is simply what identical particles are. That conclusion is false for three or more particles, and the reason is a single computable fact: the transpositions do not commute.

[\hat P_{12},\hat P_{23}]\neq 0.

where: $\hat P_{12}$ and $\hat P_{23}$ are the two pair-swaps of three particles. Because they fail to commute, they cannot be simultaneously diagonalized — there is no basis in which both have a definite eigenvalue at once, except on states where that eigenvalue is forced to agree. Concretely, $\hat P_{12}\hat P_{23}$ is a cyclic permutation of order three (verified: its cube is the identity, its square is not), the hallmark of a non-abelian group.

Figure

Swapping (1,2) then (2,3) is not the same as the reverse — the swaps do not commute

Derivation BOnly totally symmetric and totally antisymmetric states carry a definite sign for every swap

The goal is to prove that demanding a definite exchange sign for every pair simultaneously is only possible in two ways — all $+1$ or all $-1$ — and to exhibit the excluded mixed sector honestly rather than wish it away. The engine is the non-abelian structure of $S_3$ ; the payoff is the dimension census, where the excluded sector has computable, nonzero dimension that nature nonetheless never populates.

Step 1. Suppose a state $|\Psi\rangle$ is a simultaneous eigenstate of all three transpositions, $\hat P_{12}|\Psi\rangle=s_{12}|\Psi\rangle$ and so on, with each $s_{ij}=\pm 1$ . Since $\hat P_{13}=\hat P_{12}\hat P_{23}\hat P_{12}$ (conjugation within $S_3$ ), the signs are not independent: consistency forces $s_{12}=s_{23}=s_{13}$ . So the only options are all $+1$ or all $-1$ .

Step 2. But can such a simultaneous eigenstate exist at all for mixed signs? No — because $[\hat P_{12},\hat P_{23}]\neq 0$ , a generic state cannot be a joint eigenstate of both. Only the totally symmetric subspace (every $\hat P$ acts as $+1$ ) and the totally antisymmetric subspace (every $\hat P$ acts as $-1$ ) are one-dimensional representations of $S_3$ on which all group elements act by a single number. Everything else falls into higher-dimensional "mixed-symmetry" representations where the transpositions are represented by genuine matrices, not signs.

Step 3. Census the sectors. Decompose $(\mathbb C^d)^{\otimes 3}$ into $S_3$ -irreducible pieces. The totally symmetric part has dimension $\tfrac{d(d+1)(d+2)}{6}$ , the totally antisymmetric part $\tfrac{d(d-1)(d-2)}{6}$ , and the mixed part is the remainder. Computing the projector ranks numerically for $d=2,3,4,5$ reproduces these formulas exactly: for $d=3$ , $27=10+16+1$ ; for $d=4$ , $64=20+40+4$ .

Step 4. The punchline. At $d=2$ — two single-particle states, e.g. spin up and spin down — the antisymmetric dimension is $\tfrac{2\cdot 1\cdot 0}{6}=0$ . There is literally no totally antisymmetric state of three particles in a two-dimensional space: $8=4+4+0$ . Three spin- $\tfrac12$ fermions cannot all occupy two orthogonal one-particle states. That is the Pauli exclusion principle, derived here as a dimension count, before we reach the article that names it. The general $N$ -fermion constructor (the Slater determinant) is deferred by name; here we own the reason it must sometimes vanish. Why it matters: this is the operator-level answer to "why bosons and fermions and nothing else," and its $d=2$ corollary is exclusion itself. The mixed sector is not forbidden by the algebra — it is forbidden by nature, which is exactly what makes the symmetrization postulate an empirical statement rather than a theorem of quantum mechanics. Verified: the script builds the explicit $S_3$ representation, confirms $[\hat P_{12},\hat P_{23}]\neq 0$ and that $\hat P_{12}\hat P_{23}$ has order three, and computes the sym/mixed/antisym dimensions integer-exactly for $d=2$ – $5$ , including the vanishing antisymmetric dimension at $d=2$ .

The spin-exchange identity

A clean and famous corollary connects the exchange operator to spin. For two spin- $\tfrac12$ particles the swap can be written entirely in terms of the spin operators.

\hat P_{12}=\tfrac12\left(1+\boldsymbol\sigma_1\cdot\boldsymbol\sigma_2\right),

where: $\boldsymbol\sigma_i$ is the vector of Pauli matrices for particle $i$ . This is the Dirac exchange identity. Its eigenvalues sort the two-spin states cleanly: on the triplet (symmetric, spin 1) $\boldsymbol\sigma_1\cdot\boldsymbol\sigma_2=+1$ and $\hat P_{12}=+1$ ; on the singlet (antisymmetric, spin 0) $\boldsymbol\sigma_1\cdot\boldsymbol\sigma_2=-3$ and $\hat P_{12}=-1$ . The exchange operator and total-spin bookkeeping are two views of the same object — the reason spatial symmetry and spin symmetry lock together, worked out in full for the two-particle states in the next article, with the singlet/triplet construction owned by the addition-of-angular-momenta material.

Limiting cases

situation	what the algebra gives
$N=2$	one transposition; $\hat S+\hat A=\mathbb 1$ — symmetric/antisymmetric exhaust the space (deceptively complete)
$N\ge 3$	transpositions do not commute; a mixed-symmetry sector appears, and $\hat S+\hat A\neq\mathbb 1$
$d=2,\;N=3$	antisymmetric dimension $=0$ — three fermions cannot share two modes (Pauli exclusion)
label-symmetric $\hat H$	$[\hat P,\hat H]=0$ : exchange parity conserved, sectors sealed
label-asymmetric $\hat H$	not physical for identical particles; sectors would mix (control case only)
two dimensions (2D)	exchange loops braid; $S_N$ is replaced by the braid group and continuous phases (anyons) become possible

⚠️ For three particles I can just demand each pair be symmetric or antisymmetric independently

You cannot. This is the single most common misconception about many-particle symmetry, and it fails because the transpositions do not commute: $[\hat P_{12},\hat P_{23}]\neq 0$ . A state cannot carry an independent, well-defined sign for every pair — the group structure of $S_3$ forbids it unless all the signs are equal. If you try to build a "symmetric in $(1,2)$ , antisymmetric in $(2,3)$ " state, you land in the mixed-symmetry sector, where the pair-swaps are represented by genuine matrices with no single eigenvalue. Nature sidesteps the whole zoo by using only the totally symmetric and totally antisymmetric sectors. The classification of the mixed pieces is the business of Young-tableau representation theory; parastatistics is the (unrealized) hypothesis that nature might have used them.

Worked examples

Example 1 — the $S_3$ dimension census, end to end.

Step 1. Set up $d=2$ , $N=3$ . Three particles, each with a two-dimensional single-particle space (say spin up/down). Total space dimension $2^3=8$ .

Step 2. Apply the formulas. Symmetric dimension $\tfrac{d(d+1)(d+2)}{6}=\tfrac{2\cdot3\cdot4}{6}=4$ . Antisymmetric dimension $\tfrac{d(d-1)(d-2)}{6}=\tfrac{2\cdot1\cdot0}{6}=0$ . Mixed dimension $=8-4-0=4$ .

Step 3. Read the census. $8=4+4+0$ . The totally antisymmetric sector is empty: there is no way to put three spin- $\tfrac12$ particles into a fully antisymmetric state of only two modes.

Step 4. Boundary check at $d=3$ . Now $\tfrac{3\cdot4\cdot5}{6}=10$ symmetric and $\tfrac{3\cdot2\cdot1}{6}=1$ antisymmetric, so $27=10+16+1$ : the antisymmetric state reappears the moment a third mode is available (this is the $\text{det}$ of a $3\times3$ arrangement — one state). Verified: projector ranks reproduce $(4,4,0)$ and $(10,16,1)$ exactly.

Example 2 — projecting a concrete three-particle state.

Step 2. Antisymmetrize. Apply $\hat A=\tfrac1{6}\sum_P\text{sgn}(P)\hat P$ . Because the antisymmetric sector of $(\mathbb C^2)^{\otimes 3}$ is empty (Example 1), $\hat A|\Psi\rangle=\mathbf 0$ — every attempt to build a three-fermion state in two modes annihilates.

Step 3. Symmetrize. Applying $\hat S$ gives $\hat S|\Psi\rangle=\tfrac23(|001\rangle+|010\rangle+|100\rangle)$ , a genuine symmetric state that is a $+1$ eigenstate of both $\hat P_{12}$ and $\hat P_{23}$ (checked numerically).

Step 4. Norm bookkeeping. The symmetric fraction is $\dfrac{\|\hat S|\Psi\rangle\|^2}{\||\Psi\rangle\|^2}=\dfrac{4/3}{6}=\dfrac{\mathbf 2}{\mathbf 9}$ : only $22\%$ of this particular state is totally symmetric, the rest living in the mixed sector that no physical boson or fermion uses. Verified: $\hat A|\Psi\rangle=0$ , the $\hat S$ -projection is a joint $+1$ eigenstate, and the symmetric fraction is $\tfrac29$ .

The physicists

Exchange Operator

1926–1931 · Göttingen, Cambridge, Berlin

The exchange operator was born in the same months as quantum mechanics itself. In June 1926, in Göttingen, Werner Heisenberg analyzed the helium atom and discovered that the two-electron wavefunction resonates between the two labelings — the exchange degeneracy made dynamical. Two months later, in Cambridge, Paul Dirac wrote the paper that drew the full map: the many-particle wavefunction is either symmetric or antisymmetric under exchange, giving the two statistics, with no third option realized. The representation-theoretic machinery — why exactly two one-dimensional sectors survive and what the mixed pieces are — came from Eugene Wigner and Hermann Weyl between 1927 and 1931, who imported the theory of the symmetric group $S_N$ into physics. Their contemporaries, wary of the abstract group theory, called it the Gruppenpest — the "group pestilence" — a resistance that faded once the periodic table fell out of it.

⏳ History

1926

— Heisenberg analyzes helium and finds the resonance between exchanged labelings (Z. Phys. 38, 411)

1926

— Dirac establishes the symmetric/antisymmetric dichotomy for identical particles (Proc. R. Soc. A 112, 661)

1927

— Wigner brings the representation theory of the symmetric group into quantum mechanics

1928

— Weyl's group-theory book systematizes permutation and rotation symmetry for physicists

1931

— Wigner's book codifies the S_N and rotation-group methods; the "Gruppenpest" resistance recedes

Where this lives today

The exchange operator is a working piece of computational physics. In quantum chemistry and condensed-matter simulation, every many-electron wavefunction is antisymmetrized by construction (Slater determinants, and their permanents for bosons), and the projectors $\hat S,\hat A$ are the formal engine behind that. The Dirac spin-exchange identity $\hat P_{12}=\tfrac12(1+\boldsymbol\sigma_1\cdot\boldsymbol\sigma_2)$ is the origin of the Heisenberg exchange coupling that models magnetism, turning a purely kinematic symmetry into an effective spin–spin interaction. And the theorem's own frontier is alive: in two dimensions the symmetric group gives way to the braid group — double exchange stops being the identity motion, the two-sector theorem loses its premise, and a continuum of exchange phases becomes physical. Those anyons (a different escape than the never-observed mixed-symmetry parastatistics of 3D) are the subject of active experiment in fractional-quantum-Hall and cold-atom platforms.

Check yourself

Going further

Reading. Cohen-Tannoudji, Diu & Laloë, Vol. II, Ch. XIV — the permutation-operator formalism and the projector algebra in full. Sakurai & Napolitano, 2nd ed., Ch. 7 — the exchange operator and its spectrum. Griffiths, 3rd ed., Ch. 5 — the two-particle version. For the group theory of $S_N$ and Young tableaux, Hamermesh, Group Theory and Its Application to Physical Problems.

Provenance. The AllQFT base notes: "Equations and derivations cross-checked against Griffiths 3rd ed. Ch. 5, Sakurai & Napolitano 2nd ed. Ch. 7, and Cohen-Tannoudji Vol. 2 Ch. XIV. A verifier should confirm: (1) $P_{12}^2=\mathbb{1}$ forces eigenvalues $\pm1$ ; (2) the projector identities $S^2=S$ , $A^2=A$ , $SA=0$ ; (3) the spin identity $P_{12}=\tfrac12(1+\boldsymbol\sigma_1\cdot\boldsymbol\sigma_2)$ using $\boldsymbol\sigma_1\cdot\boldsymbol\sigma_2=+1$ (triplet), $-3$ (singlet); (4) that $[P_{12},H]=0$ yields conservation of exchange parity. No empirical numbers are quoted beyond standard spin eigenvalues." Every one of these four is recomputed in this article's verification script, together with the two-sector theorem and the dimension census the base did not include.

Prerequisites. Indistinguishability of Identical Particles · tensor-product Hilbert spaces · symmetries and conservation. Leads to. Symmetric and Antisymmetric States · Bosons and Fermions · Spin-Statistics Theorem · Pauli Exclusion Principle · Slater Determinants · Exchange Forces and Exchange Energy.

FAQ

:::faq[How does [P, H] = 0 square with the fact that collisions rearrange particles?] A collision changes which mode each particle occupies, but not the exchange parity of the global state. Conservation of $\langle\hat P\rangle$ says the symmetry class — symmetric or antisymmetric under label swaps — is invariant, not that individual occupations are frozen. Two colliding electrons end up in different states than they started, but the total wavefunction is antisymmetric before, during, and after. The identical-particle scattering article works out how that antisymmetry shows up as interference in the cross section.

Is the exchange operator an observable I can measure?

It is Hermitian, so formally yes — its eigenvalue is the exchange parity, $\pm 1$ . But you never build an apparatus that "measures $\hat P_{12}$ " directly; instead, the symmetrization postulate fixes the eigenvalue in advance for each species (all bosons $+1$ , all fermions $-1$ ), so the "measurement" always returns the pre-ordained value. Its real use is as a bookkeeping and projection tool, and — through the Dirac identity — as the origin of effective spin interactions.

Why can't three particles just be pairwise antisymmetric?

Because the pairwise swaps do not commute. Antisymmetry in $(1,2)$ and antisymmetry in $(2,3)$ are demands on non-commuting operators, and a state cannot be a simultaneous eigenstate of both with independent signs. The group $S_3$ only has two one-dimensional representations — the trivial (all $+1$ ) and the sign (all $-1$ ) — so "a definite sign for every pair" forces all signs equal. Anything else is a genuinely matrix-valued mixed-symmetry representation, which nature does not use for fundamental particles.

What exactly is the "mixed-symmetry" sector, and why doesn't anything live there?

It is the part of the many-particle space that transforms under higher-dimensional irreducible representations of $S_N$ — for three particles, a two-dimensional representation that the transpositions act on as $2\times 2$ matrices rather than signs. It has real, computable dimension (16 of the 27 states for three particles in three modes). No fundamental particle populates it: that restriction to the two one-dimensional sectors is the empirical content of the symmetrization postulate. The hypothesis that particles could use mixed sectors is called parastatistics, and it has never been observed.

Where does the Slater determinant come from in all this?

It is the explicit form of the antisymmetric projector $\hat A$ applied to a product of $N$ single-particle states: $\hat A(\varphi_1\otimes\cdots\otimes\varphi_N)$ equals, up to normalization, the determinant of the matrix $\varphi_i(\xi_j)$ . This article owns the projector and the two-sector theorem that guarantees such a determinant is the antisymmetric object; the general $N$ -particle determinant machinery, including when it vanishes for repeated states, is developed in the Slater determinants article. The $d=2,N=3$ vanishing here is that determinant collapsing to zero.

Does the non-commutation of transpositions change anything for two dimensions?

Profoundly. In three-dimensional space, exchanging two particles twice is topologically the same as doing nothing, which is what forces $\hat P^2=\mathbb 1$ and the $\pm 1$ eigenvalues. In two dimensions, two exchanges wind the particles around each other — a braid that cannot be undone — so $\hat P^2\neq\mathbb 1$ and the exchange can carry any phase $e^{i\theta}$ . The symmetric group $S_N$ is replaced by the braid group, and the two-sector theorem no longer applies: anyons, with a continuum of statistics, become possible. That is the subject of the anyons article.